Like Pete already said because we do this all the time we know what to
do up front. FIRST there is math done to figure out 'can we do it' in
the space available. We know to consider the difference between IBM i
6.1 and 7.1 (BIG Difference in this case!) and all of that. We know the
rules for LS migrations and what that entails such as at least one IPL
to DST and one additional IPL.
I'm not saying that doing it without tape is the best way all the time
because it's not. The closer you are to 'full' the less likely
concurrent migration will work. No doubt about that part. But we hear
horror stories about how bad things went and can see that much of it was
what we in the industry refer to as 'Self Inflicted pain.' :-)
- Larry "DrFranken" Bolhuis
On 2/28/2014 2:21 PM, rob@xxxxxxxxx wrote:
I say this with all sincerity that I know that you and a few select others
do this kind of stuff on a somewhat regular basis.
I know that it CAN be done. I just am not convinced that it IS the way to
In my case I can bring down my production machine once a quarter, over a
weekend that's planned out about a year ahead. In that case I could bang
out an unload/reload.
But when I tried the gradual, rip out x drives and replace with y bigger
drives I had to have midweek IPLs. Just not something the users expect
after coughing out big checks for Mimix, 2nd machines, etc.
Even your steps mentioned additional IPL's to DST to complete some of
Let's imagine upgrading the drives in a 5886. Let's say your running at
90+% on 70's and you want to replace them all with 140's or 280's. 90+%
is pretty darn tight to STRASPBAL to shift the load off a drive or two to
replace them 1 or 2 at a time to 140's. You end up doing quite a bit of
IPLing, etc to get through this.
Now try this on an older unit like a 595 instead of a 5886 and your raid
striping may end up a bit wonky. SAS is a bit less restrictive on Raid
rules than SCSI.